Muse Wonder: A Frontier Model For Personal & Professional AI Agents

Benchmark Model Rank Results
agents-on-jobbench-main-65-tasksMuse Wonder (max) (OpenCode)–Score: 70.5
agents-on-jobbench-main-65-tasksMuse Wonder Full (max) (OpenCode)–Score: 75.3
code-generation-on-deepsweMuse Wonder (max) (MiniSWE-Agent)–Pass@1: 0.788
code-generation-on-deepsweMuse Wonder Full (max) (MiniSWE-Agent)–Pass@1: 0.79
code-generation-on-programbenchMuse Wonder (max) (mini-SWE-agent)–Almost Resolved: 50.0
code-generation-on-programbenchMuse Wonder Full (max) (mini-SWE-agent)–Almost Resolved: 59.0
code-generation-on-terminal-bench-4-0Muse Wonder Full (max)–Accuracy: 57.1
computer-use-agents-on-osworld-2-1-108-tasksMuse Wonder (max)–Partial: 72.6Binary: 34.3
computer-use-agents-on-osworld-2-1-108-tasksMuse Wonder Full (max)–Partial: 78.4Binary: 39.4
medical-question-answering-on-healthbench-professionalMuse Wonder (max)–Score: 60.5
medical-question-answering-on-healthbench-professionalMuse Wonder Full (max)–Score: 62.3